Objective: To describe cases of tuberculosis recurrence (TBR), stratified by temporal classification (early or late TBR), and to identify possible predictors of such recurrence. Methods:This was an analytical retrospective observational epidemiological study involving a cohort of 963 new cases of pulmonary tuberculosis, reported and treated via the Tuberculosis Control Program in the city of Carapicuíba, Brazil. The study period was from 2000 to 2010. All of the pulmonary tuberculosis patients who successfully completed the treatment (with or without confirmation of cure) were selected and followed until December 31, 2012. Results: Of the 963 cases, TBR occurred in 47 (4.88%). The mean time between the first and second tuberculosis episodes was 36.12 months. Of the 47 TBR cases, 16 (34.04%) occurred within the first 18 months after the completion of the initial treatment (early TBR) and 31 (65.96%) occurred thereafter (late TBR). There were statistically significant differences between the early and late TBR groups only regarding level of education (≤ 3 vs. > 3 years of schooling; p < 0.004) and weight gain at completion of the initial treatment (1.78 kg vs. 5.31 kg; p < 0.045)-not regarding any of the other variables studied. Conclusions: A low level of education might translate to poor treatment adherence, which impedes the killing of bacilli and facilitates their survival in a latent state, making it appear as if the treatment was effective. Minimal or no weight gain at completion of the initial treatment might be a reliable biomarker to be used by health care facilities that provide tuberculosis treatment.